# plotly standard imports
import plotly.graph_objs as go
import chart_studio.plotly as py

# Cufflinks wrapper on plotly
import cufflinks

# Data science imports
import pandas as pd
import numpy as np

# Options for pandas
pd.options.display.max_columns = 30

# Display all cell outputs
from IPython.core.interactiveshell import InteractiveShell
InteractiveShell.ast_node_interactivity = 'all'

from plotly.offline import iplot, init_notebook_mode
cufflinks.go_offline(connected=True)
init_notebook_mode(connected=True)

# Set global theme
cufflinks.set_config_file(world_readable=True, theme='pearl')

Prepare dataset

from src.load_datasets import load_input_dataset

df = load_input_dataset()

df.head()

	timestamp	open	high	low	close	volume
0	1470844380000	579.00	579.00	579.00	579.00	1.000000
1	1470844440000	592.96	604.75	592.96	604.75	131.838200
2	1470844500000	604.75	604.75	600.00	600.00	89.437926
3	1470844620000	600.00	604.75	600.00	604.75	51.328200
4	1470844680000	604.75	604.75	604.75	604.75	4.586500

df.index = pd.to_datetime(df.pop('timestamp'), unit='ms')

df.head()

	open	high	low	close	volume
timestamp
2016-08-10 15:53:00	579.00	579.00	579.00	579.00	1.000000
2016-08-10 15:54:00	592.96	604.75	592.96	604.75	131.838200
2016-08-10 15:55:00	604.75	604.75	600.00	600.00	89.437926
2016-08-10 15:57:00	600.00	604.75	600.00	604.75	51.328200
2016-08-10 15:58:00	604.75	604.75	604.75	604.75	4.586500

df[::60].iplot(subplots=True)

from src.prepare_datasets import add_indicators

df = add_indicators(df)

/usr/local/lib/python3.6/dist-packages/ta/trend.py:768: RuntimeWarning:

invalid value encountered in double_scalars

/usr/local/lib/python3.6/dist-packages/ta/trend.py:772: RuntimeWarning:

invalid value encountered in double_scalars

df

	open	high	low	close	volume	volatility_bbm	volatility_bbh	volatility_bbl	trend_macd	momentum_rsi	volatility_kchi	trend_ichimoku_conv	trend_ichimoku_a	trend_ichimoku_b	momentum_stoch	momentum_stoch_signal	volatility_atr
timestamp
2016-08-10 15:53:00	579.00000	579.00	579.00000	579.00	1.000000	579.000000	579.000000	579.000000	0.000000	100.000000	0.0	579.000	579.000	579.000000	50.000000	50.000000	0.000000
2016-08-10 15:54:00	592.96000	604.75	592.96000	604.75	131.838200	591.875000	617.625000	566.125000	2.054131	100.000000	1.0	591.875	591.875	591.875000	100.000000	100.000000	0.000000
2016-08-10 15:55:00	604.75000	604.75	600.00000	600.00	89.437926	594.583333	616.960157	572.206510	3.261167	83.426791	1.0	591.875	591.875	591.875000	81.553398	90.776699	0.000000
2016-08-10 15:57:00	600.00000	604.75	600.00000	604.75	51.328200	597.125000	618.410265	575.839735	4.548603	85.936801	1.0	591.875	591.875	591.875000	100.000000	93.851133	0.000000
2016-08-10 15:58:00	604.75000	604.75	604.75000	604.75	4.586500	598.650000	618.641498	578.658502	5.505442	85.936801	1.0	591.875	591.875	591.875000	100.000000	93.851133	0.000000
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
2021-03-09 18:50:00	53725.00000	53727.00	53725.00000	53727.00	1.119537	53907.839382	54111.901593	53703.777170	-57.400039	29.446404	0.0	53820.000	53860.750	53916.339746	0.884956	0.294985	33.719104
2021-03-09 18:51:00	53780.00000	53780.00	53779.00000	53779.00	0.010000	53893.539382	54091.537996	53695.540767	-57.610695	37.411390	0.0	53820.000	53860.750	53916.339746	28.421053	9.768669	35.647194
2021-03-09 18:52:00	53814.00000	53828.00	53814.00000	53828.00	0.016450	53881.439382	54063.783077	53699.095686	-53.210375	43.844695	1.0	53820.000	53860.750	53916.339746	54.210526	27.838845	36.982474
2021-03-09 18:53:00	53903.00000	53910.00	53902.00000	53910.00	0.012500	53874.711662	54041.774570	53707.648754	-42.615134	52.621263	1.0	53820.000	53860.750	53916.339746	97.368421	60.000000	41.484227
2021-03-09 18:54:00	53930.36601	53958.00	53930.36601	53953.00	1.040000	53870.011662	54022.034074	53717.989250	-30.398177	56.463846	1.0	53841.500	53871.500	53916.339746	97.854077	83.144342	42.135804

2270370 rows × 17 columns

df[::60].iplot(subplots=True)

size = len(df)
prediction_size = 3*31*24*60

# pycarot will split on train and test by self
data = df[0:-prediction_size]
unseen_data = df[-prediction_size:]

unseen_data

	open	high	low	close	volume	volatility_bbm	volatility_bbh	volatility_bbl	trend_macd	momentum_rsi	volatility_kchi	trend_ichimoku_conv	trend_ichimoku_a	trend_ichimoku_b	momentum_stoch	momentum_stoch_signal	volatility_atr
timestamp
2020-12-06 14:50:00	19081.409767	19085.0	19081.000000	19085.0	0.076561	19106.747565	19136.215978	19077.279152	6.234778	48.197426	0.0	19089.016931	19092.258465	19034.663406	32.758621	36.206897	10.216086
2020-12-06 14:51:00	19082.000000	19082.0	19082.000000	19082.0	0.030000	19105.997565	19137.136331	19074.858799	4.640986	46.925180	0.0	19089.016931	19092.258465	19035.000000	27.586207	28.735632	9.494477
2020-12-06 14:52:00	19083.000000	19083.0	19082.000000	19082.0	0.017489	19104.073929	19136.137800	19072.010057	3.339401	46.925180	0.0	19088.993295	19092.246648	19042.500000	28.759122	29.701316	8.645030
2020-12-06 14:53:00	19082.891273	19083.0	19082.891273	19083.0	0.184115	19102.223929	19134.666123	19069.781734	2.361356	47.461316	0.0	19088.000000	19091.750000	19042.500000	30.556567	28.967299	7.880527
2020-12-06 14:54:00	19083.000000	19099.0	19083.000000	19098.0	0.099028	19101.573929	19133.806992	19069.340865	2.764753	54.831777	0.0	19085.500000	19090.500000	19045.755002	66.100406	41.805365	8.692474
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
2021-03-09 18:50:00	53725.000000	53727.0	53725.000000	53727.0	1.119537	53907.839382	54111.901593	53703.777170	-57.400039	29.446404	0.0	53820.000000	53860.750000	53916.339746	0.884956	0.294985	33.719104
2021-03-09 18:51:00	53780.000000	53780.0	53779.000000	53779.0	0.010000	53893.539382	54091.537996	53695.540767	-57.610695	37.411390	0.0	53820.000000	53860.750000	53916.339746	28.421053	9.768669	35.647194
2021-03-09 18:52:00	53814.000000	53828.0	53814.000000	53828.0	0.016450	53881.439382	54063.783077	53699.095686	-53.210375	43.844695	1.0	53820.000000	53860.750000	53916.339746	54.210526	27.838845	36.982474
2021-03-09 18:53:00	53903.000000	53910.0	53902.000000	53910.0	0.012500	53874.711662	54041.774570	53707.648754	-42.615134	52.621263	1.0	53820.000000	53860.750000	53916.339746	97.368421	60.000000	41.484227
2021-03-09 18:54:00	53930.366010	53958.0	53930.366010	53953.0	1.040000	53870.011662	54022.034074	53717.989250	-30.398177	56.463846	1.0	53841.500000	53871.500000	53916.339746	97.854077	83.144342	42.135804

133920 rows × 17 columns

from pycaret.regression import *

exp_reg_btc = setup(
    data = data, target = 'close', 
    session_id = 123, 
    log_experiment = True, log_plots=True, log_profile=True, experiment_name = 'bitcoin_close_1m',
    use_gpu = True,
    data_split_shuffle = False
)

	Description	Value
0	session_id	123
1	Target	close
2	Original Data	(2136450, 17)
3	Missing Values	False
4	Numeric Features	15
5	Categorical Features	1
6	Ordinal Features	False
7	High Cardinality Features	False
8	High Cardinality Method	None
9	Transformed Train Set	(1495514, 7)
10	Transformed Test Set	(640936, 7)
11	Shuffle Train-Test	False
12	Stratify Train-Test	False
13	Fold Generator	KFold
14	Fold Number	10
15	CPU Jobs	-1
16	Use GPU	True
17	Log Experiment	True
18	Experiment Name	bitcoin_close_1m
19	USI	abe8
20	Imputation Type	simple
21	Iterative Imputation Iteration	None
22	Numeric Imputer	mean
23	Iterative Imputation Numeric Model	None
24	Categorical Imputer	constant
25	Iterative Imputation Categorical Model	None
26	Unknown Categoricals Handling	least_frequent
27	Normalize	False
28	Normalize Method	None
29	Transformation	False
30	Transformation Method	None
31	PCA	False
32	PCA Method	None
33	PCA Components	None
34	Ignore Low Variance	False
35	Combine Rare Levels	False
36	Rare Level Threshold	None
37	Numeric Binning	False
38	Remove Outliers	False
39	Outliers Threshold	None
40	Remove Multicollinearity	False
41	Multicollinearity Threshold	None
42	Clustering	False
43	Clustering Iteration	None
44	Polynomial Features	False
45	Polynomial Degree	None
46	Trignometry Features	False
47	Polynomial Threshold	None
48	Group Features	False
49	Feature Selection	False
50	Feature Selection Method	classic
51	Features Selection Threshold	None
52	Feature Interaction	False
53	Feature Ratio	False
54	Interaction Threshold	None
55	Transform Target	False
56	Transform Target Method	box-cox

TODO

• check where MLflow store experiments data
• try put normalized data
• try use all features from add_ta_features and set feature_selection=True
• try add featues by tsfreash
• try extract features by tsfreash
• try remove trend by sktime

best = compare_models()

Initiated	. . . . . . . . . . . . . . . . . .	22:46:39
Status	. . . . . . . . . . . . . . . . . .	Compiling Final Models
Estimator	. . . . . . . . . . . . . . . . . .	Extra Trees Regressor

	Model	MAE	MSE	RMSE	R2	RMSLE	MAPE	TT (Sec)
et	Extra Trees Regressor	2194.1502	6838851.6044	2568.6732	-53.0980	0.6186	0.8305	237.7000
rf	Random Forest Regressor	2191.3906	6836494.8019	2569.2672	-53.1146	0.6191	0.8275	218.5270
xgboost	Extreme Gradient Boosting	2183.6031	6700217.1750	2543.5602	-53.2233	0.6164	0.8369	3.8400
lightgbm	Light Gradient Boosting Machine	2224.8243	6876943.7982	2575.0384	-56.9815	0.6253	0.8682	2.6250
catboost	CatBoost Regressor	2393.0245	8253141.0581	2799.5771	-58.6337	0.6477	0.9030	12.6350
gbr	Gradient Boosting Regressor	2283.0788	7160784.8496	2624.7056	-61.5425	0.6361	0.9069	303.2390
knn	K Neighbors Regressor	2367.8243	8026402.9500	2793.7685	-66.2040	0.6878	0.8956	4.3440
dt	Decision Tree Regressor	2747.6199	11521076.9317	3374.2318	-82.5318	0.8186	0.9382	17.1740
ada	AdaBoost Regressor	2355.5210	7788154.0910	2695.6864	-82.8885	0.6296	1.0305	67.8310
huber	Huber Regressor	2674.2017	11414743.0057	3161.9204	-117.9618	0.6757	1.1596	11.9880
en	Elastic Net	2658.8166	10190997.9750	2972.0072	-132.5065	0.6659	1.2163	5.2120
lasso	Lasso Regression	2658.0274	10209158.7500	2974.3060	-132.9017	0.6658	1.2169	5.6530
br	Bayesian Ridge	2657.9447	10209211.3466	2974.2967	-132.9165	0.6658	1.2169	0.3230
ridge	Ridge Regression	2657.9367	10209212.9500	2974.2957	-132.9180	0.6658	1.2169	0.0980
lr	Linear Regression	2657.9371	10209218.1250	2974.2964	-132.9181	0.6658	1.2169	0.1600
lar	Least Angle Regression	2657.9370	10209217.6058	2974.2966	-132.9181	0.6658	1.2169	0.1150
omp	Orthogonal Matching Pursuit	2687.6322	10370440.3763	2997.5908	-133.0276	0.6666	1.2234	0.1050
llar	Lasso Least Angle Regression	2999.9629	12659628.3010	3218.8327	-176.5346	0.7115	1.3961	0.1060
par	Passive Aggressive Regressor	3652.0392	21908040.9120	4451.3998	-252.5727	0.9898	1.6837	2.2180

Warning: less than 75% gpu memory available for training. Free: 4303.5625 Total: 6069.875
Warning: less than 75% gpu memory available for training. Free: 4303.4375 Total: 6069.875
Warning: less than 75% gpu memory available for training. Free: 4328.375 Total: 6069.875
Warning: less than 75% gpu memory available for training. Free: 4314.3125 Total: 6069.875
Warning: less than 75% gpu memory available for training. Free: 4321.0625 Total: 6069.875
Warning: less than 75% gpu memory available for training. Free: 4338.3125 Total: 6069.875
Warning: less than 75% gpu memory available for training. Free: 4341.625 Total: 6069.875
Warning: less than 75% gpu memory available for training. Free: 4346.5625 Total: 6069.875
Warning: less than 75% gpu memory available for training. Free: 4327.9375 Total: 6069.875
Warning: less than 75% gpu memory available for training. Free: 4334.9375 Total: 6069.875